Simulation preheating method and device for automatic driving and vehicle-mounted equipment

Abstract

The invention discloses a simulation preheating method and device for automatic driving and vehicle-mounted equipment, and the method comprises the steps: obtaining a drive test data packet of an automatic driving vehicle in a target time period, obtaining target data corresponding to a stateful function module from the drive test data packet, generating a data reduction strategy based on the state correlation and the function correlation of the stateful function module, generating state reduction data of the to-be-simulated automatic driving algorithm in the target time period according to the data reduction strategy based on the target data, and simulating the processing state of the to-be-simulated automatic driving algorithm in the target time period based on the state reduction data. According to the method, in the process of preheating the automatic driving algorithm, the data reduction strategy is generated according to the related characteristics of the stateful function module, the high reproduction success rate, the high internal state reduction rate and the low cost of the automatic driving algorithm are ensured, and the simulation success rate of the automatic driving algorithm is effectively improved through the explained cost.

Description

Technical Field

[0001] The present application relates to the technical field of simulation preheating for autonomous driving, and in particular, to a simulation preheating method, device and vehicle-mounted equipment for autonomous driving. Background Art

[0002] In response to the autonomous driving problems exposed by road tests, a logic circuit simulator (logsim) is usually used to reproduce and regress the autonomous driving algorithm. In related technologies, road tests generally use a full disk method to record all topic information, with a data size of between 1GB and 2GB per minute. After data cleaning, the input source for the simulation is obtained, which is usually a data packet containing the problem time and lasting about 30s-90s.

[0003] Simulation needs to be able to start at any time before a problem occurs. The degree of restoration of the internal state of the algorithm module is strongly correlated with the success rate of the simulation. Currently, multiple solutions for restoring the internal state of the algorithm have been tried, including cold start method, snapshot method, log open loop preheating method, and reset preheating method, etc. These algorithms are difficult to meet the requirements of high recurrence success rate, high internal state restoration rate and low cost at the same time. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the present application provides a simulation preheating method, device and vehicle-mounted equipment for autonomous driving.

[0005] The technical solution adopted by this application to solve its technical problem is:

[0006] In a first aspect, a simulation preheating method for autonomous driving is provided, comprising:

[0007] Acquire a road test data packet of an autonomous driving vehicle in a target time period, wherein the autonomous driving is implemented based on an autonomous driving algorithm to be simulated, and the autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module, wherein an output of the stateless functional module is unrelated to an input, and an output of the stateful functional module is related to an input;

[0008] Acquire target data corresponding to the stateful function module from the drive test data packet;

[0009] Generate a data restoration strategy based on the state correlation and function correlation of the stateful function modules, wherein the state correlation is the correlation between the output data and the input data of the stateful function modules, and the function correlation is the functional connection of the stateful function modules;

[0010] Based on the target data, generating state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy;

[0011] Based on the state restoration data, the processing state of the autonomous driving algorithm to be simulated in the target time period is simulated.

[0012] Furthermore, the data restoration strategy generated based on the state correlation and function correlation of the stateful functional module includes: grouping the target data to obtain multiple data groups, each data group corresponding to one of the stateful functional modules; determining the data restoration method for each of the data groups according to the state correlation and function correlation of the corresponding stateful functional module.

[0013] Furthermore, generating a data restoration strategy based on the state correlation and function correlation of the stateful function modules also includes: determining a restoration order of data by each stateful function module according to the function correlation between the stateful function modules.

[0014] Further, the at least one stateful module includes a planning module, a control module and a dynamics module, and grouping the target data includes: dividing the target data into one or more of a planning data group, a control data group and a dynamics data group, the planning data group corresponds to the planning module, the control data group corresponds to the control module, and the dynamics data group corresponds to the dynamics module.

[0015] Furthermore, if the control data group exists, grouping the target data also includes: dividing the control data group into a longitudinal data group and a transverse data group, the longitudinal data group including control data on the vehicle's driving direction, and the transverse data group including control data on vehicle steering.

[0016] Furthermore, determining the data restoration method for each data group according to the state correlation and function correlation of the corresponding stateful functional module includes: determining to adopt a first data restoration method to restore the planning data group, and the first data restoration method resets the internal variables of the planning data group; determining to adopt a second data restoration method to restore the longitudinal data group, and the second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control; determining to adopt a third data restoration method to restore the transverse data group, and the third data restoration method performs open-loop restoration processing on the data group corresponding to the transverse control; determining to adopt a fourth data restoration method to restore the dynamics data group, and the fourth data restoration method starts the dynamics module to complete the change of the control module from open-loop to closed-loop.

[0017] Furthermore, the first data restoration method resets the internal variables of the planning data group as follows: the first data restoration method resets the internal variables of the initialization information of the planning module based on a snapshot warm-up algorithm, and the initialization information includes task mode, state machine state and obstacle decision information.

[0018] Furthermore, the second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control as follows: the second data restoration method performs closed-loop restoration processing on the longitudinal data group so that the speed of the control module reaches the state at the simulation cut-in moment, including: the second data restoration method performs longitudinal preheating according to a pre-recorded real data packet so that the speed value and acceleration direction at the simulation cut-in moment are consistent with the speed value and acceleration direction in the target road test data packet.

[0019] Furthermore, the third data restoration method performs open-loop restoration processing on the data group corresponding to the lateral control as follows: the third data restoration method completes the control module while maintaining the speed and acceleration, so that the steering wheel angle and vehicle orientation inside the control reach the simulated cut-in moment state.

[0020] Furthermore, the fourth data restoration method starts the dynamics module, and completing the change of the control module from an open loop to a closed loop includes: the fourth data restoration method obtains a snapshot of vehicle status data.

[0021] Furthermore, the stateless functional module is a communication middleware, and based on the state restoration data, simulating the processing state of the autonomous driving algorithm to be simulated in the target time period includes: planning a subsequent autonomous driving trajectory based on the state restoration data and upstream autonomous driving information; running the communication middleware; based on the state restoration data, resetting the system cache data snapshot through the planning module, and resetting the vehicle state data snapshot through the dynamics module, so that the internal state of the autonomous driving algorithm to be simulated is completely consistent with the simulation cut-in moment, and the vehicle state data includes accelerator pedal data, brake pedal data, vehicle load data, vehicle gear data, steering wheel angle data, vehicle posture data, speed data and acceleration data; based on the state restoration data, controlling the vehicle driving information through the control module to be consistent with the cut-in moment, and the internal integral state is restored to the target time period.

[0022] In a second aspect, a simulation preheating device for autonomous driving is provided, comprising:

[0023] a drive test data acquisition unit, configured to acquire a drive test data packet of an autonomous driving vehicle in a target period, wherein the autonomous driving is implemented based on an autonomous driving algorithm to be simulated, and the autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module, wherein an output of the stateless functional module is unrelated to an input, and an output of the stateful functional module is related to an input;

[0024] A simulation data acquisition unit, used to acquire target data corresponding to the stateful function module from the drive test data packet;

[0025] A restoration strategy generating unit, configured to generate a data restoration strategy based on the state correlation and function correlation of the stateful function modules, wherein the state correlation is the correlation between the output data and the input data of the stateful function modules, and the function correlation is the functional connection of the stateful function modules;

[0026] A restoration data generating unit, configured to generate state restoration data of the automatic driving algorithm to be simulated in the target time period based on the target data and in accordance with the data restoration strategy;

[0027] A simulation unit is used to simulate the processing state of the autonomous driving algorithm to be simulated during the target time period based on the state restoration data.

[0028] In a third aspect, a vehicle-mounted device is provided, including:

[0029] at least one processor and at least one memory;

[0030] The memory stores executable instructions of the processor;

[0031] The processor is configured to execute the simulation preheating method for autonomous driving described in any one of the above items.

[0032] The technical solution of the present application provides a simulation preheating method, device and vehicle-mounted equipment for autonomous driving. The method obtains a road test data packet of an autonomous driving vehicle in a target time period, obtains target data corresponding to a stateful functional module from the road test data packet, generates a data restoration strategy based on the state correlation and function correlation of the stateful functional module, generates state restoration data of an autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data, and simulates the processing state of the autonomous driving algorithm to be simulated in the target time period based on the state restoration data. During the preheating process of the autonomous driving algorithm, the method generates a data restoration strategy based on relevant characteristics of the stateful functional module, thereby ensuring a high reproduction success rate, a high internal state restoration rate and a low cost of the autonomous driving algorithm, and effectively improving the simulation success rate of the autonomous driving algorithm through the cost of briefing. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 It is a flowchart of a simulation preheating method for autonomous driving provided in an embodiment of the present application;

[0035] Figure 2 It is a flowchart of another simulation preheating method for autonomous driving provided in an embodiment of the present application;

[0036] Figure 3 It is a schematic diagram of the structure of the autonomous driving algorithm module provided in the embodiment of the present application;

[0037] Figure 4 It is a schematic diagram of the execution flow of each stage in the simulation preheating method for autonomous driving provided in an embodiment of the present application;

[0038] Figure 5 It is a functional structure diagram of a simulation preheating device for autonomous driving provided in an embodiment of the present application;

[0039] Figure 6 It is a functional structure diagram of another simulation preheating device for autonomous driving provided in an embodiment of the present application;

[0040] Figure 7 It is a schematic diagram of the functional structure of the vehicle-mounted equipment provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application is described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other implementation methods obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0042] To solve this problem, refer to Figure 1 , the embodiment of the present application provides a simulation preheating method for autonomous driving, comprising:

[0043] 101. Obtain road test data packets of the autonomous driving vehicle during a target period.

[0044] The autonomous driving is implemented based on an autonomous driving algorithm to be simulated. The autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module. The output of the stateless functional module is independent of the input, and the output of the stateful functional module is dependent on the input.

[0045] 102. Acquire target data corresponding to the stateful function module from the drive test data packet.

[0046] 103. Generate a data restoration strategy based on the state correlation and function correlation of the stateful function module.

[0047] Among them, the state correlation is the correlation degree between the output data and input data of the stateful function module, and the functional correlation is the functional connection between the stateful function modules.

[0048] 104. Based on the target data, generate state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy.

[0049] 105. Based on the state restoration data, simulate the processing state of the simulated autonomous driving algorithm in the target period.

[0050] The simulation preheating method for autonomous driving provided in this embodiment obtains a road test data packet of the autonomous driving vehicle in a target time period, obtains target data corresponding to the stateful functional module from the road test data packet, generates a data restoration strategy based on the state correlation and function correlation of the stateful functional module, generates state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data, and simulates the processing state of the autonomous driving algorithm to be simulated in the target time period based on the state restoration data. In the process of preheating the autonomous driving algorithm, the method generates a data restoration strategy according to relevant characteristics of the stateful functional module, thereby ensuring a high reproduction success rate, a high internal state restoration rate and a low cost of the autonomous driving algorithm, and effectively improves the simulation success rate of the autonomous driving algorithm at the cost of disclosure.

[0051] As an improvement of the above embodiment, the embodiment of the present invention provides another simulation preheating method for autonomous driving, see Figure 2 , the method comprising:

[0052] 201. Obtain a road test data packet of the autonomous driving vehicle during a target period.

[0053] The road test data of autonomous driving refers to various data collected when the autonomous driving car is tested on the road. These data are crucial to verify the performance and safety of the autonomous driving system. Road test data mainly includes the following categories: sensor data. Autonomous driving cars are equipped with a variety of sensors, such as millimeter wave radar, laser radar and high-definition cameras. The data collected by these sensors include obstacle information, distance, speed, traffic signs, moving vehicles, pedestrians, road markings, etc. in the surrounding environment. Vehicle control and sensor status data. These data record the control status of the vehicle and the real-time data of the sensors, including the vehicle's positioning, driving speed, direction, etc. High-precision map data. Through on-board laser radar and cameras and other equipment, the driving route is three-dimensionally modeled to generate high-precision maps, providing more accurate positioning and navigation information. Safety officer takeover data. During the operation of the autonomous driving system, the data taken over by the safety officer is also recorded to analyze under what circumstances the system requires manual intervention. The collection of road test data is mainly carried out when the autonomous driving car is driving on the road. The car is equipped with various sensors and measuring equipment, such as millimeter wave radar, laser radar, high-definition camera, GPS and inertial sensor. These devices collect data about the surrounding environment and vehicle status in real time and record it through a high-performance data acquisition system. Road test data plays a vital role in the development and verification of autonomous driving technology. By collecting a large amount of real-world driving data, the perception algorithm and control logic of the autonomous driving system can be tested and optimized to improve the robustness and safety of the system. In addition, road test data can also be used to train and optimize the autonomous driving algorithm to improve the overall performance of the system.

[0054] In an embodiment of the present invention, autonomous driving is implemented based on an autonomous driving algorithm to be simulated. The autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module. The output of the stateless functional module is unrelated to the input, and the output of the stateful functional module is related to the input.

[0055] like Figure 3 As shown, the embodiment of the present invention divides the autonomous driving algorithm into a stateless module and a stateful module. The stateful functional module can be further divided into a weak state module and a strong state module. The stateless module refers to receiving upstream real-time input, the calculation process does not depend on any buffer information, the deterministic input corresponds to the deterministic output, and no preheating is required, such as communication middleware. The weak state module refers to relying on upstream real-time input and algorithm state quantities, and the output is deterministic when the internal state quantities are consistent, such as planning modules, dynamics modules, etc. The strong state module refers to relying on upstream real-time input while relying on all historical data information, and the output is deterministic when the historical data is completely consistent, such as control modules.

[0056] 202. Acquire target data corresponding to the stateful function module from the drive test data packet.

[0057] 203. Generate a data restoration strategy based on the state correlation and function correlation of the stateful function module.

[0058] Among them, the state correlation is the correlation degree between the output data and input data of the stateful function module, and the functional correlation is the functional connection between the stateful function modules.

[0059] In some optional embodiments, 203 may be implemented by, but is not limited to, the following process:

[0060] 2031. Group the target data to obtain multiple data groups, each data group corresponding to a stateful functional module.

[0061] In some optional embodiments, at least one stateful module includes a planning module, a control module and a dynamics module, and grouping the target data includes: dividing the target data into one or more of a planning data group, a control data group and a dynamics data group, the planning data group corresponds to the planning module, the control data group corresponds to the control module, and the dynamics data group corresponds to the dynamics module.

[0062] The planning module in autonomous driving is one of the core indicators for measuring and evaluating autonomous driving capabilities. Its main task is to analyze the current environment after receiving various sensory information from the sensor, and then issue instructions to the underlying control module. Typical planning modules can be divided into three levels. After receiving a given driving destination, the global route planning (Route Planning) combines map information to generate a global route as a reference for subsequent specific route planning. After receiving the global route, the behavioral decision layer (Behavioral Layer) combines environmental information obtained from the perception module (including other vehicles and pedestrians, obstacles, and traffic rules on the road) to make specific behavioral decisions (such as choosing to change lanes to overtake or follow). Finally, the motion planning (Motion Planning) layer plans and generates a trajectory that meets specific constraints (such as the vehicle's own dynamic constraints, collision avoidance, passenger comfort, etc.) based on specific behavioral decisions. This trajectory is used as the input of the control module to determine the final driving path of the vehicle.

[0063] The control module in the autonomous driving system is an important part of the autonomous driving system. It is mainly responsible for accurately controlling the vehicle to drive along the planned trajectory according to the instructions of the decision-making and planning module. The control module generates specific acceleration, steering and braking instructions to control the drive system, steering system, braking system and suspension system to ensure that the vehicle runs smoothly on the established trajectory. The control module generates control instructions. According to the instructions output by the decision-making and planning module, the control module generates specific acceleration, steering and braking instructions to ensure that the vehicle drives along the planned trajectory. The control module also monitors the vehicle's status and environmental changes in real time, and makes adjustments as needed to ensure the stability and safety of the vehicle's driving. The control module also interacts with the actuator: The control module interacts with the vehicle's actuators (such as the drive system, steering system, braking system and suspension system) through communication protocols such as the CAN bus to implement specific control actions. The control module is closely related to the perception module and the planning module. The perception module is responsible for collecting environmental information around the vehicle, the planning module calculates the vehicle's driving path and actions based on the perception information, and the control module generates specific control signals based on these instructions to ensure that the vehicle drives along the planned trajectory.

[0064] The dynamics module in autonomous driving is a module used to process and analyze the dynamic characteristics of the vehicle, mainly involving the mechanical behavior and dynamic response of the vehicle during motion. In the autonomous driving system, the dynamics module is an important part of motion control. It ensures the stability and safety of the vehicle in various driving scenarios by analyzing the dynamic characteristics of the vehicle. The dynamics module first needs to establish the dynamic model of the vehicle, including the motion equations of the vehicle in the global coordinate system and the road coordinate system. These models take into account factors such as the lateral force, longitudinal force, and yaw rate of the vehicle, and reflect the mechanical behavior of the vehicle under different driving conditions. The dynamics module is also used for dynamic response analysis. Through the dynamics model, the module can analyze the dynamic response of the vehicle under different driving conditions, such as high-speed cornering, emergency braking, etc. This helps to ensure the stability and safety of the vehicle under these extreme conditions. The dynamics module is used for controller design. Based on the dynamics model, a suitable controller can be designed to adjust the driving state of the vehicle, such as adjusting the throttle, brake, and steering wheel, to achieve the desired trajectory and speed. In the autonomous driving system, the dynamics module interacts closely with other modules such as the environmental perception module, the planning module, and the control module. Specifically: The environmental perception module provides environmental information around the vehicle, such as obstacle location, speed, etc. The planning module plans the driving trajectory and speed according to environmental information and traffic rules. The motion control module realizes precise control of the vehicle through dynamic model analysis and controller design based on the planned trajectory and speed.

[0065] Further, in some optional embodiments, if a control data group exists, grouping the target data also includes: dividing the control data group into a longitudinal data group and a transverse data group, the longitudinal data group including control data on the vehicle's driving direction, and the transverse data group including control data on the vehicle's steering.

[0066] 2032. Determine the data restoration method for each data group according to the state correlation and function correlation of the corresponding stateful function module.

[0067] In some optional embodiments, 2032 may be implemented by, but not limited to, the following process (not shown in the figure):

[0068] 2032-1. Determine to use a first data restoration method to restore the planning data group, and reset internal variables of the planning data group using the first data restoration method.

[0069] In some optional embodiments, the first data restoration method resets the internal variables of the planning data group as follows: the first data restoration method resets the internal variables of the initialization information of the planning module based on a snapshot warm-up algorithm, and the initialization information includes the task mode, state machine state and obstacle decision information.

[0070] 2032-2. Determine to use a second data restoration method to restore the longitudinal data group, and perform closed-loop restoration processing on the data group corresponding to the longitudinal control by the second data restoration method.

[0071] In some optional embodiments, the second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control: the second data restoration method performs closed-loop restoration processing on the longitudinal data group so that the control module speed reaches the state at the simulation cut-in moment, including: the second data restoration method performs longitudinal preheating according to a pre-recorded real data packet, so that the speed value and acceleration direction at the simulation cut-in moment are consistent with the speed value and acceleration direction in the target road test data packet.

[0072] 2032-3. Determine to use a third data restoration method to restore the horizontal data group. The third data restoration method performs open-loop restoration processing on the data group corresponding to the horizontal control.

[0073] In some optional embodiments, the third data restoration method performs open-loop restoration processing on the data group corresponding to the lateral control: the third data restoration method completes the control module while maintaining the speed and acceleration, so that the steering wheel angle and vehicle direction inside the control reach the simulated cut-in moment state.

[0074] 2032-4. Determine to use the fourth data restoration method to restore the kinetic data group. Use the fourth data restoration method to start the kinetic module and complete the change of the control module from open loop to closed loop.

[0075] The embodiment of the present invention divides the simulated autonomous driving algorithm warm-up scheme into stages, namely, a state initialization stage, a longitudinal warm-up stage, a dynamic warm-up stage, a lateral warm-up stage and a simulation stage. The state initialization stage and the dynamic warm-up stage can be adjusted in time sequence as needed. Figure 4 One possible execution order is shown.

[0076] The above 2032-1 corresponds to the state initialization stage. This stage is based on the snapshot method. The internal variables of the weak state module are reset. The initialization information of the planning module is the internal buffer snapshot of the algorithm at the time of simulation entry, including but not limited to the task mode, state machine state, obstacle decision information, etc. After entering the simulation stage, the subsequent trajectory is planned based on the reset state and upstream information. The snapshot method refers to recording the algorithm state information as a snapshot, and using the interface to inject it back to the algorithm module when starting.

[0077] The above 2032-2 corresponds to the longitudinal preheating stage, which is mainly for the strong state module, so that the control module speed reaches the state at the simulation cut-in moment. This stage belongs to open-loop preheating, and the specific steps are as follows: pre-record the real data packet, the vehicle moves forward, accelerates from the stationary moment to the maximum speed, moves at a constant speed, decelerates to the stationary state, and reverses in the same way; the simulation module selects the appropriate preheating package according to the simulation cut-in moment, and injects it into the control module in an open-loop manner, controls the expected speed and the speed at the cut-in moment to be within the critical threshold, and the acceleration direction is consistent, and the longitudinal preheating is completed.

[0078] The above 2032-4 corresponds to the dynamics warm-up stage, which completes the change from open loop to closed loop and starts the dynamics model. The initial information of the dynamics module is the snapshot of the vehicle state at the time of simulation entry, including but not limited to the accelerator pedal, brake pedal, vehicle load, vehicle gear, steering wheel angle, vehicle posture, speed, acceleration, etc. After this stage, the prefabricated package data is stopped, and the dynamics receives the control command and feedbacks the vehicle state.

[0079] The above 2032-3 corresponds to the lateral preheating stage, which is mainly for the strong state module. The control module maintains the speed and acceleration while making the steering wheel angle and vehicle orientation inside the control reach the state at the time of simulation cut-in. This stage belongs to the dynamic closed-loop stage. The specific steps are as follows: the dynamic model is initialized based on the vehicle snapshot, receives the real-time instructions issued by the control, calculates the actual state of the vehicle, and before feedback, modifies the vehicle position to the position at the time of simulation cut-in to ensure that the internal position of the control does not change, modifies the actual vehicle speed to be consistent with the control expected speed, ensures that the longitudinal integral of the control remains unchanged, maintains the longitudinal state, and truthfully feeds back the wheel angle and vehicle posture. When the control steering wheel angle expectation and the steering wheel angle at the time of cut-in are within the critical threshold, reset the dynamic model snapshot, and the lateral preheating is completed.

[0080] In some optional embodiments, the fourth data restoration method starts the dynamics module, and completing the change of the control module from an open loop to a closed loop includes: the fourth data restoration method obtains a snapshot of the vehicle status data.

[0081] Furthermore, in some optional embodiments, 203 may also include:

[0082] 2033. Determine the order in which each stateful functional module restores data according to the functional correlation between the stateful functional modules.

[0083] 204. Based on the target data, generate state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy.

[0084] 205. Based on the state restoration data, the processing state of the simulated autonomous driving algorithm in the target time period is simulated.

[0085] After all modules are preheated, they enter the simulation phase. In this phase, the stateless module (communication middleware) operates normally, the weak-state module planning system resets the buffer snapshot, the dynamics model resets the vehicle snapshot, the internal state is completely consistent with the simulation entry moment, the strong-state module controls the internal speed, vehicle posture, position and other information to be consistent with the entry moment (within the threshold), the internal integral state is restored, there is no integral mutation, the protection system will not be accidentally triggered to cause braking or even stopping, the connection is completed smoothly, and the vehicle starts with the state.

[0086] In some optional embodiments, the stateless functional module is a communication middleware, and 205 can be implemented by, but not limited to, the following process (not shown in the figure):

[0087] 2051. Plan subsequent autonomous driving trajectories based on state restoration data and upstream autonomous driving information.

[0088] 2052. Run the communication middleware.

[0089] 2053. Based on the state restoration data, the system cache data snapshot is reset through the planning module, and the vehicle state data snapshot is reset through the dynamics module, so that the internal state of the autonomous driving algorithm to be simulated is completely consistent with the simulation entry moment. The vehicle state data includes accelerator pedal data, brake pedal data, vehicle load data, vehicle gear data, steering wheel angle data, vehicle posture data, speed data and acceleration data.

[0090] 2054. Based on the state restoration data, the vehicle driving information is controlled by the control module to be consistent with the cut-in time, and the internal integral state is restored to the target period.

[0091] The simulation preheating method for autonomous driving provided in this embodiment obtains the road test data packet of the autonomous driving vehicle in the target period, obtains the target data corresponding to the stateful function module from the road test data packet, generates a data restoration strategy based on the state correlation and function correlation of the stateful function module, generates the state restoration data of the autonomous driving algorithm to be simulated in the target period according to the data restoration strategy based on the target data, and simulates the processing state of the autonomous driving algorithm to be simulated in the target period based on the state restoration data. In the process of preheating the autonomous driving algorithm, the method generates a data restoration strategy according to the relevant characteristics of the stateful function module to ensure the high recurrence success rate, high internal state restoration rate and low cost of the autonomous driving algorithm, and effectively improves the simulation success rate of the autonomous driving algorithm through the cost of the disclosure. The problem of low success rate of the existing related preheating method is solved. Taking the open-pit mine case as a reference, the recurrence and regression success rates are both above 90%. The recurrence success rate is equivalent to the log open source preheating method, while the time consumption is reduced by 80%; the regression success rate is increased from 40%-50% to above 90%.

[0092] In summary, the simulation preheating method for autonomous driving provided by the embodiment of the present invention is universal. Users can use this method to reset the state of any road test data, including but not limited to forward, reverse, curve, ramp and stationary. It has been verified in practice that the success rate is >= 90%. Without secondary adaptation, users can use any algorithm version for regression with a success rate of >= 90%. The method is efficient, and the preheating time of any road test data is within controllable limits. It can quickly complete the preheating and reduce time costs. The method also has high reducibility and efficiency. It has been developed and put into use. According to production environment statistics, under the premise of ensuring that the reproduction success rate is greater than 90%, the preheating time of a single case is shortened from 20-30 minutes to 0-2 minutes, which can save more than 80% of time. The algorithm version is changed for backtesting, and the simulation success rate is increased from 30-40% to 90%+.

[0093] In order to cooperate with the above-mentioned automatic driving simulation preheating method, the embodiment of the present invention provides an automatic driving simulation preheating device, see Figure 5 ,include:

[0094] The road test data acquisition unit 51 is used to obtain the road test data packet of the autonomous driving vehicle in the target time period. The autonomous driving is implemented based on the autonomous driving algorithm to be simulated. The autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module. The output of the stateless functional module is not related to the input, and the output of the stateful functional module is related to the input.

[0095] The simulation data acquisition unit 52 is used to acquire target data corresponding to the stateful function module from the drive test data packet.

[0096] The restoration strategy generating unit 53 is used to generate a data restoration strategy based on the state correlation and function correlation of the stateful function modules. The state correlation is the correlation between the output data and the input data of the stateful function modules. The function correlation is the functional connection of the stateful function modules.

[0097] The restored data generating unit 54 is used to generate the state restored data of the automatic driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data.

[0098] The simulation unit 55 is used to simulate the processing state of the simulated automatic driving algorithm in the target time period based on the state restoration data.

[0099] The simulation preheating device for autonomous driving provided in the present embodiment obtains a road test data packet of the autonomous driving vehicle in a target time period, obtains target data corresponding to the stateful functional module from the road test data packet, generates a data restoration strategy based on the state correlation and function correlation of the stateful functional module, generates state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data, and simulates the processing state of the autonomous driving algorithm to be simulated in the target time period based on the state restoration data. In the process of preheating the autonomous driving algorithm, the method generates a data restoration strategy according to relevant characteristics of the stateful functional module, thereby ensuring a high reproduction success rate, a high internal state restoration rate and a low cost of the autonomous driving algorithm, and effectively improves the simulation success rate of the autonomous driving algorithm at the cost of disclosure.

[0100] As an improvement of the above embodiment, the embodiment of the present invention provides another simulation preheating device for automatic driving, see Figure 6 ,include:

[0101] The road test data acquisition unit 61 is used to obtain the road test data packet of the autonomous driving vehicle in the target time period. The autonomous driving is implemented based on the autonomous driving algorithm to be simulated. The autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module. The output of the stateless functional module is not related to the input, and the output of the stateful functional module is related to the input.

[0102] The simulation data acquisition unit 62 is used to acquire target data corresponding to the stateful function module from the drive test data packet.

[0103] The restoration strategy generating unit 63 is used to generate a data restoration strategy based on the state correlation and function correlation of the stateful function modules. The state correlation is the correlation between the output data and the input data of the stateful function modules. The function correlation is the functional connection of the stateful function modules.

[0104] In some optional embodiments, the restoration strategy generation unit 63 includes:

[0105] The data grouping module 631 is used to group the target data into multiple data groups, each data group corresponds to a stateful function module.

[0106] In some optional embodiments, at least one stateful module includes a planning module, a control module and a dynamics module, and the data grouping module 631 groups the target data including: the data grouping module 631 divides the target data into one or more of a planning data group, a control data group and a dynamics data group, the planning data group corresponds to the planning module, the control data group corresponds to the control module, and the dynamics data group corresponds to the dynamics module.

[0107] In some optional embodiments, if a control data group exists, the data grouping module 631 further includes grouping the target data: the data grouping module 631 divides the control data group into a longitudinal data group and a transverse data group, the longitudinal data group includes control data on the vehicle's driving direction, and the transverse data group includes control data on the vehicle's steering.

[0108] The data group restoration method determination module 632 is used to determine the data restoration method of each data group according to the state correlation and function correlation of the corresponding stateful function module.

[0109] In some optional embodiments, the data group restoration method determination module 632 includes:

[0110] The first data restoration submodule 6321 is used to determine whether to use the first data restoration method to restore the planning data group. The first data restoration method resets the internal variables of the planning data group. Specifically, the first data restoration method resets the internal variables of the planning data group as follows: the first data restoration method resets the internal variables of the initialization information of the planning module based on a snapshot preheating algorithm. The initialization information includes the task mode, state machine state and obstacle decision information.

[0111] The second data restoration submodule 6322 is used to determine whether to use the second data restoration method to restore the longitudinal data group. The second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control. Specifically, the second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control as follows: the second data restoration method performs closed-loop restoration processing on the longitudinal data group so that the speed of the control module reaches the state at the simulation cut-in moment, including: the second data restoration method performs longitudinal preheating according to a pre-recorded real data packet, so that the speed value and acceleration direction at the simulation cut-in moment are consistent with the speed value and acceleration direction in the target road test data packet.

[0112] The third data restoration submodule 6323 is used to determine the third data restoration method for data restoration of the lateral data group. The third data restoration method performs open-loop restoration processing on the data group corresponding to the lateral control. Specifically, the third data restoration method performs open-loop restoration processing on the data group corresponding to the lateral control as follows: the third data restoration method completes the control module while maintaining the speed and acceleration, so that the steering wheel angle and vehicle orientation inside the control reach the simulated cut-in moment state.

[0113] The fourth data restoration submodule 6324 is used to determine to use the fourth data restoration method to restore the dynamics data group, and the fourth data restoration method starts the dynamics module to complete the change of the control module from open loop to closed loop. Specifically, the fourth data restoration method starts the dynamics module to complete the change of the control module from open loop to closed loop, including: the fourth data restoration method obtains a snapshot of the vehicle state data.

[0114] The data restoration sequence determination module 633 is used to determine the data restoration sequence of each stateful functional module according to the functional correlation between the stateful functional modules.

[0115] The restored data generating unit 64 is used to generate the state restored data of the automatic driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data.

[0116] The simulation unit 65 is used to simulate the processing state of the simulated autonomous driving algorithm in the target time period based on the state restoration data.

[0117] In some optional embodiments, the stateless functional module is a communication middleware, and the simulation unit 65 includes:

[0118] The trajectory planning module 651 is used to plan the subsequent autonomous driving trajectory based on the state restoration data and the upstream autonomous driving information.

[0119] The communication middleware running module 652 is used to run the communication middleware.

[0120] The data synchronization module 653 is used to restore the data based on the state. It resets the system cache data snapshot through the planning module and the vehicle state data snapshot through the dynamics module, so that the internal state of the autonomous driving algorithm to be simulated is completely consistent with the simulation entry time. The vehicle state data includes accelerator pedal data, brake pedal data, vehicle load data, vehicle gear data, steering wheel angle data, vehicle posture data, speed data and acceleration data.

[0121] The vehicle information synchronization module 654 is used to restore the data based on the state, control the vehicle driving information to be consistent with the cut-in time through the control module, and restore the internal integration state to the target period.

[0122] The simulation preheating device for autonomous driving provided in the present embodiment obtains a road test data packet of the autonomous driving vehicle in a target time period, obtains target data corresponding to the stateful functional module from the road test data packet, generates a data restoration strategy based on the state correlation and function correlation of the stateful functional module, generates state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data, and simulates the processing state of the autonomous driving algorithm to be simulated in the target time period based on the state restoration data. In the process of preheating the autonomous driving algorithm, the method generates a data restoration strategy according to relevant characteristics of the stateful functional module, thereby ensuring a high reproduction success rate, a high internal state restoration rate and a low cost of the autonomous driving algorithm, and effectively improves the simulation success rate of the autonomous driving algorithm at the cost of disclosure.

[0123] Based on the same inventive concept, Figure 7 As shown, the present application also provides a vehicle-mounted device, including:

[0124] at least one processor 71 and at least one memory 72;

[0125] The memory stores executable instructions for the processor;

[0126] The processor is configured to execute the simulation preheating method for autonomous driving provided in the above-mentioned embodiment.

[0127] The simulation preheating system for autonomous driving provided in the embodiment of the present application stores executable instructions of the processor in a memory. When the executable instructions are executed, the processor can obtain a road test data packet of the autonomous driving vehicle in a target time period, obtain target data corresponding to the stateful functional module from the road test data packet, generate a data restoration strategy based on the state correlation and function correlation of the stateful functional module, generate state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy based on the target data, simulate the processing state of the autonomous driving algorithm to be simulated in the target time period based on the state restoration data, and generate a data restoration strategy according to the relevant characteristics of the stateful functional module during the preheating process of the autonomous driving algorithm, thereby ensuring a high reproduction success rate, a high internal state restoration rate and a low cost of the autonomous driving algorithm, and effectively improving the simulation success rate of the autonomous driving algorithm through the cost of briefing.

[0128] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0129] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

Claims

1. A simulation preheating method for autonomous driving, characterized in that: include: Acquire a road test data packet of an autonomous driving vehicle in a target time period, wherein the autonomous driving is implemented based on an autonomous driving algorithm to be simulated, and the autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module, wherein an output of the stateless functional module is unrelated to an input, and an output of the stateful functional module is related to an input; Acquire target data corresponding to the stateful function module from the drive test data packet; Generate a data restoration strategy based on the state correlation and function correlation of the stateful function modules, wherein the state correlation is the correlation between the output data and the input data of the stateful function modules, and the function correlation is the functional connection of the stateful function modules; Based on the target data, generating state restoration data of the autonomous driving algorithm to be simulated in the target time period according to the data restoration strategy; Based on the state restoration data, the processing state of the autonomous driving algorithm to be simulated in the target time period is simulated.

2. The method according to claim 1, characterized in that: The generating of the data restoration strategy based on the state correlation and function correlation of the stateful function module comprises: Grouping the target data to obtain a plurality of data groups, each data group corresponding to one of the stateful function modules; The data restoration method of each data group is determined according to the state correlation and function correlation of the corresponding stateful function modules.

3. The method according to claim 2, characterized in that: The generating of the data restoration strategy based on the state correlation and function correlation of the stateful function module further comprises: The order in which the data are restored by each stateful functional module is determined according to the functional correlation between the stateful functional modules.

4. The method according to claim 3, characterized in that: The at least one stateful module includes a planning module, a control module, and a dynamics module, and grouping the target data includes: The target data is divided into one or more of a planning data group, a control data group and a dynamics data group, the planning data group corresponds to the planning module, the control data group corresponds to the control module, and the dynamics data group corresponds to the dynamics module.

5. The method according to claim 4, characterized in that: If the control data group exists, grouping the target data further includes: The control data set is divided into a longitudinal data set and a transverse data set, wherein the longitudinal data set includes control data in the vehicle's driving direction and the transverse data set includes control data on the vehicle's steering direction.

6. The method according to claim 5, characterized in that: Determining the data restoration method of each data group according to the state correlation and function correlation of the corresponding stateful function module includes: Determining to use a first data restoration method to restore the planning data group, wherein the first data restoration method resets internal variables of the planning data group; Determining to use a second data restoration method to restore the longitudinal data group, wherein the second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control; Determining to use a third data restoration method to restore the horizontal data group, wherein the third data restoration method performs an open-loop restoration process on the data group corresponding to the horizontal control; A fourth data restoration method is determined to be used for data restoration of the kinetic data group, and the fourth data restoration method starts the kinetic module to complete the change of the control module from an open loop to a closed loop.

7. The method according to claim 6, characterized in that: The first data restoration method resets the internal variables of the planning data group as follows: the first data restoration method resets the internal variables of the initialization information of the planning module based on a snapshot warm-up algorithm, and the initialization information includes the task mode, state machine state and obstacle decision information.

8. The method according to claim 7, characterized in that: The second data restoration method performs closed-loop restoration processing on the data group corresponding to the longitudinal control as follows: the second data restoration method performs closed-loop restoration processing on the longitudinal data group so that the speed of the control module reaches the state at the simulation cut-in moment, including: the second data restoration method performs longitudinal preheating according to a pre-recorded real data packet so that the speed value and acceleration direction at the simulation cut-in moment are consistent with the speed value and acceleration direction in the target road test data packet.

9. The method according to claim 8, characterized in that: The third data restoration method performs open-loop restoration processing on the data group corresponding to the lateral control as follows: the third data restoration method completes the control module while maintaining the speed and acceleration, so that the steering wheel angle and vehicle orientation inside the control reach the simulated cut-in moment state.

10. The method according to claim 9, characterized in that: The fourth data restoration method starts the dynamics module and completes the change of the control module from an open loop to a closed loop, including: the fourth data restoration method obtains a snapshot of vehicle status data.

11. The method according to claim 10, characterized in that: The stateless function module is a communication middleware, and based on the state restoration data, simulating the processing state of the to-be-simulated autonomous driving algorithm in the target time period includes: planning a subsequent autonomous driving trajectory based on the state restoration data and upstream autonomous driving information; Running the communication middleware; Based on the state restoration data, the planning module resets the system cache data snapshot, and the dynamics module resets the vehicle state data snapshot, so that the internal state of the autonomous driving algorithm to be simulated is completely consistent with the simulation cut-in time, and the vehicle state data includes accelerator pedal data, brake pedal data, vehicle load data, vehicle gear data, steering wheel angle data, vehicle posture data, speed data and acceleration data; Based on the state restoration data, the vehicle driving information is controlled by the control module to be consistent with the cut-in time, and the internal integral state is restored to the target period.

12. A simulation preheating device for automatic driving, characterized in that: include: a drive test data acquisition unit, configured to acquire a drive test data packet of an autonomous driving vehicle in a target period, wherein the autonomous driving is implemented based on an autonomous driving algorithm to be simulated, and the autonomous driving algorithm to be simulated is implemented by a stateless functional module and at least one stateful functional module, wherein an output of the stateless functional module is unrelated to an input, and an output of the stateful functional module is related to an input; A simulation data acquisition unit, used to acquire target data corresponding to the stateful function module from the drive test data packet; A restoration strategy generating unit, configured to generate a data restoration strategy based on the state correlation and function correlation of the stateful function modules, wherein the state correlation is the correlation between the output data and the input data of the stateful function modules, and the function correlation is the functional connection of the stateful function modules; A restoration data generating unit, configured to generate state restoration data of the automatic driving algorithm to be simulated in the target time period based on the target data and in accordance with the data restoration strategy; A simulation unit is used to simulate the processing state of the autonomous driving algorithm to be simulated during the target time period based on the state restoration data.

13. A vehicle-mounted device, characterized in that: include: at least one processor and at least one memory; The memory stores executable instructions of the processor; The processor is configured to execute the simulation preheating method for autonomous driving as described in any one of claims 1-11.

Images